The expression profile of GSE32323, deposited by Khamas et al (10) was downloaded from the Gene Expression Omnibus (www.ncbi.nlm.nih.gov/geo/), which was based on the platform of the GPL570 (HG-U133_Plus_2) Affymetrix Human Genome U133 Plus 2.0 array (Affymetrix, Inc., Santa Clara, CA, USA). The GSE32323 profile included a collection of 17 pairs of matched cancer-normal colorectal tissue samples. The CRC samples were obtained from the American Type Culture Collection (Manassas, VA, USA) and the Cell Resource Center for Biomedical Research of Tohoku University (Sendai, Japan). The cells were cultured in media obtained from Sigma-Aldrich; Merck Millipore (Darmstadt, Germany) or Gibco; Thermo Fisher Scientific, Inc. (Waltham, MA, USA), supplemented with 100 µg/ml streptomycin (Invitrogen; Thermo Fisher Scientific, Inc.), 100 U/ml penicillin and 10% heat-inactivated fetal bovine serum (Nichirei Biosciences, Tokyo, Japan).

Following downloading of the GSE32323 profile, the microarray data were preprocessed using the Affy package (11) in Bioconductor (www.ncbi.nlm.nih.gov/geo/) and the Affy probe annotation file from Brain Array Lab (brainarray.mhri.med.umich.edu/brainarray/default.asp). In brief, the process included background correction, quantile normalization, probe summarization, and log2 conversion. The mean values of probes mapped with the same gene were obtained as ultimate gene expression values. Subsequently, the Linear Models for Microarray Data (Limma) package in R (www.bioconductor.org/packages/release/bioc/html/limma.html) (12) was used to screen the DEGs of the CRC samples, compared with the normal samples. An adjusted P-value of P<0.05 and |logfold-change|>1 were used as the cut-off criteria. The CRC genes were identified from a search using the Online Mendelian Inheritance in Man (OMIM) database (www.ncbi.nlm.nih.gov/omim) (13) and were then annotated as seed genes.

The PPI network was downloaded from the Search Tool for the Retrieval of Interacting Genes database (string-db.org/) (14). The sub-network of the DEGs and seed genes was then obtained and termed the CRC.PPI. The RWR algorithm was originally used for image segmentation (15). With a restarting probability (r) of 0.9, the R package dnet (16) was used to perform RWR analysis to identify the top 50 nodes with the highest affinity scores in the CRC.PPI. The affinity score referred to the proximity between two nodes and the parameter ‘r’ is the probability of moving to seed nodes.

The Database for Annotation, Visualization and Integrated Discovery (DAVID; david.abcc.ncifcrf.gov/) is a tool for agglomerating a large quantity of gene/protein symbols into gene clusters (17). Gene Ontology (GO) analysis is used to predict the potential functions of large-scale transcriptomic data or genomic data (18). The Kyoto Encyclopedia of Genes and Genomes database is a base for functional analysis combing with networks of genes and other molecules (19). Using the DAVID online tool, GO functional and pathway enrichment analyses were performed for the DEGs in the PPI network of the top 50 nodes. P<0.05 was used as the cut-off criterion.

The Drug-Gene Interaction database (dgidb.genome.wustl.edu/) (20) was used to identify interactions between antineoplastic drugs and genes. The actions of the drugs inhibit genes expressed at high levels and promote genes expressed at low levels.

Compared with the normal samples, there were 1,640 DEGs screened in the CRC samples, including 850 upregulated genes and 790 downregulated genes. The CRC genes or seed genes from the OMIM database are listed in Table I, including cyclin D1 (CCND1) and aurora kinase A (AURKA).

Following downloading of the PPI network and obtaining the CRC.PPI of the DEGs and seed genes, RWR analysis was performed and nodes with the top 50 affinity scores, which included DEP domain-containing MTOR-interacting protein (DEPTOR), breast carcinoma amplified sequence-1 (BCAS1), neural precursor cell expressed developmentally downregulated 9 (NEDD9), mitogen-activated protein kinase kinase 2, (MAP2K2) in the CRC.PPI were identified. The PPI network of the top 50 nodes with the highest affinity scores, which comprised 14 upregulated genes, 17 downregulated genes and 19 non-DEGs, had 224 interactions. These interactions included DEPTOR-CCND1, AURKA-BCAS1, CCND1-BCAS1, CCND1-NEDD9 and CCND1-MAP2K2 (Fig. 1). The DEGs included in the top 50 nodes are listed in Table II. The enriched functions of the 31 DEGs in the PPI network of the top 50 nodes were primarily associated with cell cycle, including cell cycle phase (P=0.005651), M phase (P=0.019763) and intracellular signaling cascade (P=0.019784), as shown in Table III. The enriched pathway for the 31 DEGs in the PPI network of the top 50 nodes included cytokine-cytokine receptor interaction (P=0.04602; Table III).

The interactions between antineoplastic drugs and the 31 DEGs in the CRC.PPI network were screened. Only three DEGs (CCND1, AURKA and DEPTOR), had interactions with their corresponding antineoplastic drugs.

For the analysis of drug-gene interactions in the present study, a total of 1,640 DEGs were screened in the CRC samples, compared with normal samples, which included 850 upregulated genes and 790 downregulated genes. The obtained seed genes included CCND1 and AURKA. The enriched functions for the 31 DEGs in the PPI network of the top 50 nodes were primarily associated with cell cycle. According to the drug-gene interaction analysis, only three DEGs (CCND1, AURKA and DEPTOR) had interactions with their corresponding antineoplastic drugs.

DEPTOR, which is negatively correlated with tumor progression and the activity of mammalian target of rapamycin complex 1 (mTORC1) in CRC, may function as a marker for prognosis and treatment of CRC (21). The PI3K/Akt/mTOR signaling pathway is one of the primary mechanisms involved in maintaining tumor progression and metastasis, and mTOR activation/inhibition functions in CRC (22). It has been reported that DEPTOR may be closely correlated with CRC. The variants in CCND1 may have potential associations with familial CRC (23), and the CCND1 A870G polymorphism may increase the risk of CRC (24). The epidermal growth factor A61G and CCND1 A870G polymorphisms may be valuable molecular biomarkers for the prognosis of patients with CRC treated with the single-agent, cetuximab (25). Thus, the expression levels of CCND1 may be associated with CRC. The PPI network of the top 50 nodes in the present study also showed CCND1 had an interaction with DEPTOR, indicating that CCND1 may be involved in CRC by mediating DEPTOR.

Independent of established clinicopathological variables, AURKA can act as a molecular marker for predicting clinical outcome in patients with liver metastasis of CRC (26). AURKA, which is located on chromosome 20q, is involved in adenoma-to-carcinoma progression and may function as an indicator of poor prognosis (27–33). NABC1, also known as BCAS1, and its variant, NABC15B, are downregulated in CRC (34). These findings suggest that AURKA and BCAS1 may be associated with CRC. In the PPI network of the top 50 nodes in the present study, it was also found that BCAS1 interacted with AURKA, suggesting that BCAS1 may also be involved in CRC through regulating AURKA.

As a scaffolding protein, NEDD9, also termed human enhancer of filamentation 1 (HEF1) is involved in mediating several cellular processes, including cell cycle progression, cellular attachment, apoptosis, motility and inflammation (35,36). Prostaglandin E2 (PGE2) promotes the expression of HEF1 and then induces cell cycle progression, and HEF1 is an important downstream mediator of the activity of PGE2 in the progression of CRC (37). The activation and overexpression of extracellular signal-regulated kinase/MAPK is important in the progression of CRC and can be a molecular target for the treatment of the cancer (38). It has been reported that MAP2K inhibitors can be of use in CRC therapy (39). Therefore, the dysregulation of NEDD9 and MAP2K2 may affect CRC. In the PPI network of the top 50 nodes in the present study, CCND1 was also found to have interactions with BCAS1, NEDD9 and MAP2K2, indicating that CCND1 may also function in CRC through BCAS1, NEDD9 and MAP2K2.

In conclusion, the present study performed a comprehensive bioinformatics analysis of genes, which may be involved in CRC. A total of 1,640 DEGs were screened in the CRC samples, compared with normal samples. It was found that DEPTOR, AURKA, CCND1, BCAS1, NEDD9 and MAP2K2 may be correlated with CRC. However, further confirmation is required to elucidate their mechanisms of action in CRC.